Trace Element Levels in Drinking Water and Cognitive Function

American Journal of Epidemiology
Copyright © 2000 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 151, No. 9
Printed in U.S.A.
Trace Element Levels in Drinking Water and Cognitive Function among
Elderly Chinese
Christine L. Emsley,1 Sujuan Gao,1 Yiming Li, z Chaoke Liang,3 Rongdi Ji, 3 Kathleen S. Hall," Jingxiang Cao,3
Feng Ma,3 Yunpeng Wu,3 Po Ying,3 Yan Zhang,3 Shuzhuang Sun,3 Frederick W. Unverzagt,4 Charles W.
Slemenda,1* and Hugh C. Hendrie4
aged; cadmium; calcium; cognition; trace elements; water supply; zinc
In recent years, a number of studies have examined
the relations between trace elements, cognitive function, and/or Alzheimer's disease. Many of these studies have involved measuring trace elements in drinking
water, and most of them have focused on the effect of
aluminum on cognitive function. The evidence that
increased levels of aluminum in drinking water are
associated with Alzheimer's disease is inconclusive at
best (1, 2). In addition to aluminum, other elements
such as calcium, iron, zinc, and fluoride have been
measured. Jacqmin et al. (3, 4) reported a positive
association between calcium levels in drinking water
and cognitive function but no association with iron and
zinc. Still and Kelley (5) found lower rates of hospital
admission for primary degenerative dementia (primar-
ily Alzheimer's disease) in a South Carolina county
with high fluoride levels as compared with two low
fluoride counties.
Other studies have examined the effect of minerals
on cognitive function by measuring dietary intake or
blood levels. Ortega et al. (6) reported a positive relation between cognitive function and dietary intakes of
zinc and iron in healthy elderly adults. In a study of
blood levels of iron and zinc, Tucker et al. (7) reported
that three indices of iron status (plasma iron, transferrin, and ferritin) varied in their relation to cognitive
performance and that plasma zinc was not significantly
correlated with cognitive performance. It has been
suggested that antioxidants such as vitamin E and selenium might have a protective effect on the development of Alzheimer's disease, although findings have
been mixed. A study of antioxidant supplement use
(including zinc and selenium) showed, after adjustment for age, that there was no significant difference in
performance on cognitive tests between antioxidant
users and nonusers (8). However, a recent analysis of
data collected in the Third National Health and
Nutrition Examination Survey did show a positive
relation between vitamin E blood levels and cognitive
function (9). Blood lead levels, even within low levels
of lead exposure, are negatively correlated with cognitive function in elderly men (10) and women (11).
Studies assessing nutritional status through either
dietary intake or blood levels can provide very detailed
Received for publication January 29,1999, and accepted for publication July 7, 1999.
Abbreviations: CSI"D", Community Screening Interview for
Dementia; EPA, Environmental Protection Agency.
1
Department of Medicine, Indiana University School of Medicine,
Indianapolis, IN.
2
Department of Restorative Dentistry, Loma Linda University
School of Dentistry, Loma Linda, CA.
3
Institute of Environmental Health and Engineering, Chinese
Academy of Preventive Medicine, Beijing, China.
4
Department of Psychiatry, Indiana University School of Medicine, Indianapolis, IN.
* Deceased.
Reprint requests to Christine L. Emsley, Division of Biostatistics,
Indiana University School of Medicine, 1001 West Tenth Street,
RG/4th, Indianapolis, IN 46202.
913
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
The relation between trace element levels in drinking water and cognitive function was investigated in a
population-based study of elderly residents (n = 1,016) in rural China in 1996-1997. Cognitive function was
measured using a Chinese translation of the Community Screening Interview for Dementia. A mixed effects
model was used to evaluate the effect of each of the elements on cognitive function while adjusting for age, sex,
and educational level. Several of the elements examined had a significant effect on cognitive function when they
were assessed in a univariate context. However, after adjustment for other elements, many of these results were
not significant. There was a significant quadratic effect for calcium and a significant zinc-cadmium interaction.
Cognitive function increased with calcium level up to a certain point and then decreased as calcium continued
to increase. Zinc showed a positive relation with cognitive function at low cadmium levels but a negative relation
at high levels. Am J Epidemiol 2000; 151:913-20.
914
Emsley et al.
MATERIALS AND METHODS
Subjects
The study was carried out among two populations in
the Henan and Shandong provinces of northern China
in 1996-1997 as part of a study on long term fluoride
exposure and bone fracture rates. Water sources within
the villages in these regions were identified, and several trace elements and pH levels were measured. A
complete census of all subjects over age 65 years was
carried out in each village, and all residents were asked
to participate in the study. A total of 1,024 subjects
agreed to participate. Subjects were not systematically
excluded for comorbid conditions, except for eight
subjects who had extreme hearing or sight impairment
which made completion of the cognitive assessment
unfeasible. There were no refusals.
Trace element levels
Samples of drinking water were collected at the
study sites and analyzed for nine elements: fluoride,
calcium, selenium, aluminum, iron, zinc, cadmium,
lead, and arsenic. The fluoride level of water was
determined directly using an ion-selective electrode
(Orion Research, Inc., Beverly, Massachusetts). The
methods used for analysis of the other eight elements
followed Chinese National Standard procedures,
which were set by the Standard Department of the
Chinese Academy of Preventive Medicine (19). Water
calcium and selenium levels were determined using
the ethylenediaminetetraacetic acid titrimetric method
and the fluorescence method, respectively, and a spectrophotometric procedure with acetic acid and ammonium acetate was used for the evaluation of aluminum
in drinking water. For determination of water iron and
zinc levels, flame atomic absorption spectrophotometry was employed, while levels of water cadmium and
lead were analyzed using the graphite furnace atomic
absorption spectrophotometry technique. The concentration of arsenic in drinking water was determined by
means of the silver diethyldithiocarbamate method.
Cognitive function
Cognitive function was assessed using a slightly
modified Chinese translation of the Community
Screening Interview for Dementia (CSI"D"). The
CSI"D" is an instrument that was originally developed
and validated in a study comparing the Cree Indians in
Manitoba, Canada, with Manitobans of European
descent (20). It has now been used with a high degree
of reliability among Yoruba, African Americans, and
Jamaicans (21). The instrument has two parts: a cognitive and risk factor section for the subject and an interview with a relative about the daily functioning and
general health of the subject. For this analysis, we used
data from the cognitive assessment only. The cognitive
function assessment includes items from widely used
dementia instruments, including the Cambridge
Mental Disorders of the Elderly Examination (22), the
Mini-Mental State Examination (23), the Dementia
Rating Scale (24), and the Comprehensive Assessment
and Referral Evaluation (25). The CSI"D" was specifically designed for use in populations with both literate and illiterate elderly persons, such as this one. A
Am J Epidemiol
Vol. 151, No. 9, 2000
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
and accurate information about current nutritional status, but they do not provide information on lifetime
exposure to trace minerals.
There have also been some speculative articles on
the possible relation between trace elements and the
neuropathologic changes associated with Alzheimer's
disease. Deibel et al. (12) reported elevated zinc and
iron levels in the brains of patients who died of
Alzheimer's disease as compared with normal brains,
while Tully et al. (13) showed a moderate to strong
negative correlation between fasting serum zinc concentrations and senile plaque counts in seven brain
regions. Khachaturian (14) has proposed a calcium
hypothesis which states that the cellular mechanisms
for maintaining the homeostasis of cellular calcium
concentration play a key role in aging.
Many trace elements have been reported to interact
with each other in a biologic context. For example,
lead impairs normal calcium homeostasis in cells, and
low dietary calcium may increase lead levels in critical
organs (15). Experimentally, lead has been shown to
increase zinc excretion (15). Optimal iron intake can
protect against lead toxicity (16). Cadmium can interact with calcium, zinc, and iron (15). Zinc can reduce
iron uptake (16). Whanger (17) reported a possible
selenium-fluoride interaction. However, most previous
studies have failed to take these interactions into
account.
Studying the relation between trace elements in drinking water and cognitive function in many areas of the
world is very difficult, if not impossible, for several reasons. Many people are very mobile and tend to change
residences frequently throughout their lifetime. They
also consume foods and beverages that were prepared
and packaged in different areas of the world. In contrast,
residents of rural China rarely move, and most are
exposed to the same water supply throughout their lifetime. Their food and drinks come from local sources.
For these reasons, rural China has been considered a
perfect "living laboratory" for studying relations
between various environmental factors and diseases
(18).
Trace Elements and Cognitive Function
full description of the development of the original
instrument and its reliability and validity has been published elsewhere (26).
The cognitive assessment consists of 30 items aimed
at assessing cognitive function in three areas: memory,
language, and attention. In addition, there is a short
story recall with 14 possible items and a naming fluency question which asks subjects to name as many
foods as they can in 1 minute. The cognitive score was
calculated as a simple sum of all items, including a
point for each detail recalled from the story and a point
for each food named in the fluency question. Possible
scores range from zero to 44, plus the number of foods
listed in the fluency question.
TABLE 1. Demographic data and mean cognitive scores of
rural elderly Chinese, 1996-1997
Males (n = 444)
Females {n = 572)
Mean
age in
years
Education
(% with
no
schooling)
72.1 (5.1)*
72.0 (5.5)
57.7
96.9*
Mean
cognitive
scoret
44.5 (9.6)*
38.5 (8.8)*
* Significant difference between males and females (p < 0.05).
t Score on the Community Screening Interview for Dementia
(20,21).
* Numbers in parentheses, standard deviation.
Demographic data for males versus females were
compared using t tests for continuous variables and
chi-squared tests for categorical variables. Spearman's
correlation coefficients were calculated to examine the
relations among elements. Mixed effects models were
used to analyze the relation between cognitive function
and trace elements. Cognitive score was the outcome
variable and age, sex, and educational level (no education vs. some education) were included in the model as
covariates. Water source was included as a random
effect to account for the correlation between subjects
sharing the same water source.
We analyzed the effects of the various trace elements
in a three-step process. First, we examined each element
individually in a mixed model for its effect on cognitive
function (univariate analysis). Each element was tested
for both a linear and a quadratic effect. Next, we tested
interactions for pairs of elements which were previously
reported to affect each other in a biologic context
(bivariate analysis). Our third and final step was to fur-
RESULTS
Cognitive scores for the 1,016 subjects included in
the study ranged from 12 to 77, with a mean score of 41
and a standard deviation of 9.6. The median score was
42. Basic demographic data and mean cognitive scores
are presented by gender in table 1. Males and females
did not differ significantly according to age. A larger
percentage of females than of males had no schooling
(p < 0.01), and males scored significantly higher on the
cognitive function assessment (p < 0.01). Table 2
shows the distribution of trace element levels in the 20
water sources included in the study. The limits set by
the US Environmental Protection Agency (EPA) (27,
28) and the National Sanitary Standards for Drinking
Water in China (19) are also presented for reference.
Note that there are values above the EPA limits for iron
and fluoride, but all other values are within the limits.
Levels of trace elements and pH in 20 water sources assessed in rural northern China, 1996-1997
Element
Mean
(SDt)
Minimum
Median
Maximum
Chinese limit*
US limitU
Cadmium
(ng/liter)
Calcium
(mg/liter)
Fluoride
(mg/liter)
Iron
(ng/liter)
Lead
(ng/liter)
Selenium
(ng/liter)
Zinc
(jig/liter)
Zinc*
(Mg/liter)
pH
0.22
(0.112)
0.07
0.25
0.49
10.0
5.0
72.2
(33.27)
30.2
59.8
153.10
No limit§
No limit
2.6
(0.94)
1.45
2.7
4.70
1.0
4.0
266.7
(307.8)
15.4
165.4
1,330
300
300#
2.2
(2.69)
0.29
0.29
9.46
50.0
15.0
0.53
(0.946)
0.05
0.12
4.24
10
50
15.21
(52.26)
0.15
0.15
236
1,000
5,000#
3.6
(5.63)
0.15
0.15
17.9
1,000
5,000#
7.4
(0.20)
7.1
7.4
7.8
6.5-8.5
6.5-8.5#
* Distribution of zinc values with outlier removed,
t SD, standard deviation.
* Chinese National Sanitary Standards for Drinking Water in China (19).
§ Water hardness limit: calcium carbonate concentration of 450 mg/liter (19).
1) Environmental Protection Agency (EPA) limit (27).
* Secondary EPA limit; recommended, but not required (28).
Am J Epidemiol Vol. 151, No. 9, 2000
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
ther examine any interaction or main effect that was significant in the bivariate or univariate analyses. For each
significant effect (univariate or bivariate), we tested the
effect while adjusting for any other element that was
significantly correlated with it (multivariate analysis).
Statistical analyses
TABLE 2.
915
916
Emsleyetal.
Univariate analyses
The results from the analysis of individual elements
are presented in table 3. Each element was tested for
both a linear effect and a quadratic effect. Calcium had
TABLE 3. Results from mixed effects models testing the
effects of trace element levels and pH on cognitive function in
elderly Chinese subjects, 1996-1997*
Linear model
Cadmium (ng/liter)
Calcium (mg/liter)
Fluoride (mg/liter)
Iron (ng/liter)
Lead (ng/liter)
PH
Selenium (ng/liter)
Zinc (ng/liter)
Quadratic model
Estimate
P
value
Estimate
P
value
-0.356
-0.0078
0.899
0.0039
-0.191
1.29
-0.656
-0.130
0.93
0.48
0.02
0.01
0.08
0.42
<0.01
0.01
38.3
-0.0009
-0.424
0.00
0.072
11.5
0.404
0.002
0.20
<0.01
0.32
0.20
0.03
0.14
0.07
0.87
* Results were adjusted for age, sex, and educational level.
Parameter estimates represent the change in cognitive score
(Community Screening Interview for Dementia (20, 21)) for each
unit increase in trace element level.
a significant negative quadratic effect (p < 0.01) on
cognitive function after adjustment for age, sex, and
education. This means that cognitive function
increased as calcium level increased up to a point, and
then cognitive function decreased as calcium level
continued to increase. Lead had a significant positive
quadratic effect {p = 0.03). As lead level increased,
cognitive function decreased up to a point; then cognitive function increased as lead level continued to
increase. Fluoride and iron both showed a significant
positive linear relation with cognitive function {p =
0.02 and p = 0.01, respectively). Selenium and zinc
each demonstrated a significant negative linear effect
(p < 0.01 and/? = 0.01, respectively).
Bivariate analyses
After conducting the univariate analysis, we examined pairs of elements which were previously reported
to interact with each other in a biologic context. A statistically significant interaction would indicate that the
effect of one element changed depending on the level
of another element. Results are presented in table 4.
Interactions between cadmium and calcium, cadmium
and iron, lead and calcium, lead and zinc, fluoride and
selenium, and iron and zinc were not significant at the
0.05 level. There was a significant interaction between
cadmium and zinc (p — 0.02) after adjustment for age,
sex, and educational level. At low levels of cadmium,
zinc showed a positive relation with cognitive function, while at high levels of cadmium, zinc showed a
negative relation. There was also a significant leadiron interaction {p = 0.02). For low levels of iron, lead
had a negative relation with cognitive function, while
at high levels of iron, lead had no significant effect on
cognitive function.
Am J Epidemiol
Vol. 151, No. 9, 2000
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
The results of analysis of the 20 water sources
showed that eight of the sources had an aluminum
level of 0.01 mg/liter and the other 12 sources had a
level of 0.025 mg/liter. These levels are well below the
EPA recommended limits. Similarly, there were only
three distinct arsenic levels. Twelve sites had an
arsenic level of 0.005 mg/liter, six had a level of 0.010
mg/liter, and two had a level of 0.020 mg/liter. These
levels are also within the limits of the EPA water regulations. Because we had only a minimal range of levels for these two elements, we could not analyze their
effect on cognitive function. We can assume that
within these analyses we controlled for aluminum and
arsenic.
The various trace element levels were distributed
relatively evenly across the range of the data, except
in the case of zinc, iron, and selenium. These three
elements had one water source each which had an
extremely high value, relative to the rest of the data.
All analyses for these three elements were run both
with and without the most extreme value included.
In the cases of iron and selenium, the exclusion of
the extreme value made no difference in the results
and conclusions, so we chose not to exclude the
extreme values. When the water source with a zinc
level of 236 (Xg/liter was excluded from the analyses
involving zinc, we obtained different results.
Although this value was a valid data point, we chose
to exclude it from the analysis. Because we had no
data between 17.9 |lg/liter and 236.0 (Xg/liter, there
was no way to make valid conclusions about the
effect of zinc between those levels. There were 123
subjects who shared the water source with the high
zinc level, and they were excluded from any analysis
involving zinc.
Spearman's correlation coefficients among trace elements and pH were calculated for examination of the
relations among the elements. Selenium was correlated
positively with lead (r = 0.86, p < 0.01) and zinc (r =
0.63, p < 0.01) but negatively with fluoride (r = 0.81,
p < 0.01) and iron (r = -0.61, p < 0.01). Fluoride was
positively correlated with iron (r = 0.57, p < 0.01) and
negatively correlated with zinc (r = -0.65, p < 0.01)
and lead (r = -0.80, p < 0.01). Zinc was positively
correlated with lead (r = 0.73, p < 0.01). Iron was negatively correlated with lead (r = -0.48, p < 0.03) and
positively correlated with pH (r = 0.60, p < 0.01).
None of the other correlations were significant at the
0.05 level.
Trace Elements and Cognitive Function
917
TABLE 4. Results from mixed effects models testing the effects of interactions between trace elements on cognitive function in
elderly Chinese subjects, 1996-1997*
Element 1
Element 1
Element 2
Cadmium (ng/liter)
Cadmium (jxg/Iiter)
Cadmium (ng/liter)
Lead (ng/liter)
Lead (ng/liter)
Lead (jxg/Iiter)
Fluoride (mg/liter)
Iron (ng/liter)
Calcium (jig/liter)
Iron (ng/liter)
Zinc (fig/liter)
Calcium (mg/liter)
Zinc (ng/liter)
Iron (ng/liter)
Selenium (fig/liter)
Zinc (fxg/liter)
Element 2
Interaction
Estimate
p
value
Estimate
P
value
Estimate
P
value
7.78
6.01
6.09
-0.09
-0.35
-0.54
0.04
0.001
0.60
0.34
0.12
0.83
0.18
0.89
0.94
0.73
0.02
0.72
0.05
0.03
0.99
0.68
0.02
0.09
0.02
-1.22
-0.01
-6.66
-0.002
-0.019
0.003
4.07
0.001
0.60
0.31
0.02
0.82
0.60
0.02
0.11
0.10
0.007
1.53
-0.0001
0.087
0.0003
-7.57
-O.20
Multivariate analyses
Because several of the elements measured in this
study were strongly correlated with each other, additional analyses were carried out for effects that were
significant in the univariate and bivariate analyses: fluoride, selenium, cadmium and zinc, and lead and iron.
Although calcium had a significant effect, it was not
significantly correlated with any of the other elements.
This analysis was conducted in order to verify, to the
extent we could, that any significant effect we were
seeing was actually the effect of that element and was
not due simply to the presence or absence of some
other element. We did not include all elements in one
large model, because the distribution of the element
levels arising from the observational nature of the
study would probably cause such a model to be statistically unstable. For example, to fit such a model, we
would need sites with ranges of selenium at each level
of cadmium, calcium, fluoride, iron, lead, zinc, and pH
and at each combination of these elements. We would
need such ranges for all combinations of all elements
included in the model, which is difficult to achieve in
an observational study.
Selenium was significantly correlated with iron,
lead, zinc, and fluoride, so we examined the effect of
selenium while adjusting for each of these elements.
Results are presented in table 5, sections A, B, and C.
The effect of selenium was not significant after adjustment for lead and iron (p — 0.61), cadmium and zinc
(p = 0.45), and fluoride (p = 0.06). These results support the conclusion that selenium does not have a significant effect on cognitive function, although the
effect was significant in the univariate analysis.
Fluoride was correlated with iron, lead, selenium,
and zinc. Results from the adjusted analysis for fluoride are shown in table 5, sections C, D, and E. As with
selenium, the effect of fluoride was not significant
when the analysis was adjusted for these other eleAm J Epidemiol Vol. 151, No. 9, 2000
ments. Similarly, we conclude that fluoride is not significantly related to cognitive function.
Iron was correlated with fluoride and pH, and lead
was correlated with fluoride, selenium, and zinc. The
corresponding analyses are presented in sections A, D,
F, and G of table 5. Although the iron-lead interaction
remained significant after adjustment for pH, it was
not significant after adjustment for the other elements
that were correlated with either iron or lead.
Cadmium was not significantly correlated with any
of the other elements, but zinc was correlated with fluoride, lead, and selenium. The appropriate adjusted
analyses are shown in parts B, E, and F of table 5.
Unlike the fluoride and selenium effects, the zinccadmium interaction remained significant even after
we adjusted for the elements that were correlated with
zinc.
DISCUSSION
In this Chinese population, we found that calcium
had a quadratic effect on cognitive function. Jacqmin
et al. had previously reported a positive relation
between calcium and cognitive function (3,4). We also
found a positive relation, but only up to calcium levels
of approximately 86 mg/liter. At calcium levels greater
than 86 mg/liter, the relation reversed; cognitive function decreased as calcium level increased. This relation
is displayed graphically in figure 1. We have not found
any evidence that calcium levels above 86 mg/liter in
drinking water are toxic. However, calcium is closely
related to water hardness, and it may be an indicator of
concentration of elements not measured in this study.
We also found a significant zinc-cadmium interaction, even after adjusting for other elements. Although
an interaction between zinc and cadmium had been
previously reported (15), it had not been reported in
the context of cognitive function. We saw that at low
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
* Results were adjusted for age, sex, and educational level. Parameter estimates represent the change in cognitive score (Community
Screening Interview for Dementia (20, 21)) for each unit increase in trace element level.
918
Emsleyetal.
TABLE 5. Results from multivariate mixed effects models
testing the effects of trace elements on cognitive function in
elderly Chinese subjects, 1996-1997*
P
value
Selenium (jag/liter)
-0.17
Lead ((ig/liter)
Iron (u.g/liter)
Lead x iron
-0.43
0.0006
0.002
0.61
0.18
0.80
0.14
B. Selenium, adjusted for zinc
and cadmium
Selenium (|ig/liter)
Zinc (|ig/liter)
Cadmium (ng/liter)
Zinc x cadmium
-O.30
1.49
6.11
-6.31
0.45
0.03
0.12
0.03
C. Selenium, adjusted for fluoride
Selenium (u.g/liter)
Fluoride (mg/liter)
-0.50
0.49
0.06
0.25
D. Fluoride, adjusted for iron and
lead
Fluoride (mg/liter)
Lead (ng/liter)
Iron (|ig/liter)
Lead x iron
0.02
-0.54
0.0003
0.003
0.97
0.05
0.90
0.03
E. Fluoride, adjusted for zinc and
cadmium
Fluoride (mg/liter)
Zinc ((ig/liter)
Cadmium (u.g/liter)
Zinc x cadmium
0.49
1.49
5.88
-6.36
0.30
0.03
0.13
0.02
F. Zinc and cadmium, adjusted for
lead and iron
Zinc (ng/liter)
Cadmium (ng/liter)
Zinc x cadmium
Lead (u.g/liter)
Iron (|ig/liter)
Lead x iron
1.71
8.44
-6.85
-0.77
-O.003
0.003
0.04
0.06
0.04
0.05
0.29
0.18
-0.59
0.001
0.003
-1.50
0.01
0.68
0.01
0.39
A. Selenium, adjusted for lead
and iron
G. Lead and iron, adjusted for pH
Lead (ng/liter)
Iron ((j.g/liter)
Lead x iron
PH
* Results were adjusted for age, sex, and educational level.
Parameter estimates represent the change in cognitive score
(Community Screening Interview for Dementia (20, 21)) for each
unit increase in trace element level.
levels of cadmium, zinc had a positive relation with
cognitive function. This agrees with results reported
by Ortega et al. (6). However, when cadmium was at
high levels, zinc showed a negative relation with cognitive function. Figure 2 demonstrates this relation
graphically by plotting the least squares means for
Am J Epidemiol
Vol. 151, No. 9, 2000
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
Estimate
Variable(s)
cognitive score (adjusted for age, sex, and education)
for groups based on cadmium level and zinc level.
In the univariate analyses, calcium, fluoride, iron,
lead, selenium, and zinc all had significant effects on
cognitive function. However, after adjustment for
other elements and possible interactions, calcium was
the only effect which remained significant. This highlights the importance of measuring multiple trace elements simultaneously, as the relation between naturally occurring levels of elements and their effect on
cognition is probably very complex.
The analysis of trace elements in drinking water and
their effect on cognitive function poses several difficulties. Some of these difficulties are resolved by
studying a group such as rural Chinese residents. The
subjects included in this study were unique in that
many of them had lived in the same village for their
entire lives. They also consumed almost exclusively
food and beverages that were grown and prepared
locally. In the areas selected for the study, we performed a complete census of people over age 65 rather
than just taking a sample. We also had data on nine different trace elements, seven of which we analyzed, and
pH. Because of this, we were able to examine not only
the effects of individual elements but also their interactions.
Although this study of trace elements in drinking
water overcame many of the obstacles that other studies
of this type face, there are still a few difficulties in
determining the relation between trace element levels
and cognitive function. We must consider the fact that
levels of elements in drinking water cannot provide allinclusive information on exposure levels. Subjects can
also be exposed to trace elements through their diet and
through the air. However, an extensive Chinese study of
selenium and fluoride showed that element levels in
water, air, and blood were highly correlated (29). On the
basis of that study, we made the assumption that trace
element levels in drinking water would give information indicative of personal exposure, even though it is
not all-inclusive information. This assumption, for elements other than selenium and fluoride, must be verified by future studies that have data on all exposure
sources. We also assumed that these Chinese residents
had lived under fairly constant environmental conditions throughout their lifetimes—specifically, that they
had had the same water source throughout their lives.
This may not have been true for female subjects. It is
traditional for Chinese women to move to their husbands' villages after marriage. As a result, their exposure levels would have been constant over the past
40-50 years rather than over their entire lifetimes.
However, the impact of marriage on the data collected
from the females may have been insignificant. In these
Trace Elements and Cognitive Function
919
80
100
Calcium Level
FIGURE 1. Relation between calcium intake and cognitive function in rural China, 1996-1997. The plot was generated by a weighted local
regression model regressing mean cognitive score on calcium level. The weight was the number of subjects used to calculate each mean score.
rural populations, particularly 40-50 years ago, marriage often occurred among residents of the same village or nearby villages that had substantial similarity in
geographic conditions, water sources, food, and culture.
Nevertheless, in the present study, we included gender
in all of our models to control for this possible effect.
The complex relations among naturally occurring levels of elements introduce difficulties into the statistical
analysis of this type of data. Our multivariate analysis of
the data involved the inclusion of elements that were
correlated with each other in the same model, which can
introduce problems with multicollinearity. Because
most of our parameter estimates did not change drastically when additional variables were added, and also
because our standard errors were small relative to the
parameter estimates, we did not feel that multicollinearity was a serious problem for the models presented here.
The investigation of trace elements in drinking
water as possible risk factors for cognitive impairment
and Alzheimer's disease is potentially important,
because the chemical composition of drinking water is
modifiable, while many other risk factors are not (30).
This population of rural Chinese provided us with an
opportunity to study trace element exposures over a
lifetime. The study sites were selected so that they
would provide a range of water fluoride levels. For
future studies, we propose to select sites that will allow
us to evaluate the association between cognitive performance and such elements as aluminum, and to
study a greater range of elements (such as selenium).
—
_
-
—
-
-
"
43-
42-
Low Cadmium
High Cadmium
High
Zinc Level
FIGURE 2. Relation between the interaction of zinc and cadmium intakes and cognitive function in rural China, 1996-1997. Results were generated from a mixed effects model adjusted for age, sex, and educational level. Zinc and cadmium were split at the median exposure level for
illustration purposes.
Am J Epidemiol
Vol. 151, No. 9, 2000
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
8-
920
Emsley et al.
ACKNOWLEDGMENTS
This work was supported in part by National Institute of
Aging grant P30 AG10133 and a grant from Eli Lilly and
Company (Indianapolis, Indiana).
REFERENCES
Am J Epidemiol Vol. 151, No. 9, 2000
Downloaded from http://aje.oxfordjournals.org/ at Indiania University School of Dentistry on December 28, 2011
1. Munoz DG. Is exposure to aluminum a risk factor for the
development of Alzheimer disease?—No. Arch Neural 1998;
55:737-9.
2. Forbes WF, Hill GB. Is exposure to aluminum a risk factor for
the development of Alzheimer disease?—Yes. Arch Neurol
1998;55:740-l.
3. Jacqmin H, Commenges D, Letenneur L, et al. Components of
drinking water and risk of cognitive impairment in the elderly.
Am J Epidemiol 1994;139:48-57.
4. Jacqmin-Gadda H, Commenges D, Letenneur L, et al. Silica
and aluminum in drinking water and cognitive impairment in
the elderly. Epidemiology 1996;7:281-5.
5. Still CN, Kelley P. On the incidence of primary degenerative
dementia vs. water fluoride content in South Carolina.
Neurotoxicology 1980;4:125-31.
6. Ortega RM, Requejo AM, Andres P, et al. Dietary intake and
cognitive function in a group of elderly people. Am J Clin Nutr
1997;66:803-9.
7. Tucker DM, Penland JG, Sandstead HH, et al. Nutrition status
and brain function in aging. Am J Clin Nutr 1990;52:93-102.
8. Mendelsohn AB, Belle SH, Stoehr GP, et al. Use of antioxidant supplements and its association with cognitive function
in a rural elderly cohort: the MoVIES Project. Monongahela
Valley Independent Elders Survey. Am J Epidemiol 1998; 148:
38^14.
9. Perkins AJ, Hendrie HC, Callahan CM, et al. Association of
antioxidants with memory in a multiethnic elderly sample
using the Third National Health and Nutrition Examination
Survey. Am J Epidemiol 1999; 150:37-44.
10. Payton M, Riggs KM, Spiro A III, et al. Relations of bone and
blood lead to cognitive function: the VA Normative Aging
Study. Neurotoxicol Teratol 1998;20:19-27.
11. Muldoon SB, Cauley JA, Kuller LH, et al. Effects of blood
lead levels on cognitive function of older women. Neuroepidemiology 1996;15:62-72.
12. Deibel MA, Ehmann WD, Markesbery WR. Copper, iron, and
zinc imbalances in severely degenerated brain regions in
Alzheimer's disease: possible relation to oxidative stress. J
Neurol Sci 1996; 143:137^2.
13. Tully CL, Snowdon DA, Markesbery WR. Serum zinc, senile
plaques, and neurofibrillary tangles: findings from the Nun
Study. Neuroreport 1995;6:2105-8.
14. Khachaturian ZS. Calcium hypothesis of Alzheimer's disease
and brain aging. Ann N Y Acad Sci 1994;747:1-11.
15. Goyer RA. Toxic and essential metal interactions. Annu Rev
Nutr 1997;17:37-50.
16. Berdanier CD. Trace elements—iron. In: Advanced nutrition
micronutrients. Boca Raton, FL: CRC Press, 1998:187-93.
17. Whanger PD. China, a country with both selenium deficiency
and toxicity: some thoughts and impressions. I Nutr 1989; 119:
1236-9.
18. Moffat AS. China: a living lab for epidemiology. Science
1990;248:553-5.
19. Standard Department, Chinese Academy of Preventive
Medicine. Standard compilation of environment, school and
radiation health. Beijing, China: Chinese Standard Publishing
House, 1991:278-338.
20. Hall KS, Hendrie HC, Rodgers D, et al. The development of a
dementia screening interview in two distinct languages. Int J
Methods Psychiatr Res 1993;3:l-28.
21. Hall KS, Oyediran O, Li Y, et al. Development and performance of an assessment of cognitive function for crosscultural studies. (Abstract). Neurobiol Aging 1998;19(suppl):
S170.
22. Roth M, Tym E, Mountjoy CQ, et al. CAMDEX: a standardised instrument for the diagnosis of mental disorder in the
elderly with special reference to the early detection of dementia. Br J Psychiatry 1986; 149:698-709.
23. Folstein MF, Folstein SE, McHugh PR. "Mini-Mental State":
a practical method for grading the cognitive state of patients
for the clinician. J Psychiatr Res 1975; 12:189-98.
24. Blessed G, Tomlinson BE, Roth M. The association between
quantitative measures of dementia and of senile change in the
cerebral grey matter of elderly subjects. Br I Psychiatry 1968;
• 114:797-811.
25. Gurland B, Kuriansky J, Sharpe L, et al. The Comprehensive
Assessment and Referral Evaluation (CARE): rationale, development and reliability. Int J Aging Hum Dev 1977;8:9-42.
26. Hall KS, Ogunniyi AO, Hendrie HC, et al. A cross-cultural
community based study of dementias: methods and performance of the survey instrument—Indianapolis, U.S.A. and
Ibadan, Nigeria. Int J Methods Psychiatr Res 1996;6:129^2.
27. Environmental Protection Agency. National primary drinking
water regulations. Code of Federal Regulations, volume 40,
part 141, July 1, 1997.
28. Environmental Protection Agency. National secondary drinking water regulations. Code of Federal Regulations, volume
40, part 143, July 1, 1997.
29. Cao J, Yan B, Zhang S, et al. Heavy exposure to selenium-fluoride and human health. (In Chinese). J Hyg Res 1995;25:
287-90.
30. Forbes WF, Gentleman JF, Maxwell CJ. Concerning the role of
aluminum in causing dementia. Exp Gerontol 1995;30:23-32.